METHODOLOGY · v1.0.0
How the AI Ready Score is computed.
The score is deterministic: the same inputs against the same data snapshot always produce the same number. No language model touches the math. Every result carries its methodology version and data versions, so any score can be reproduced and audited.
THE FORMULA
One score, four pillars.
Your overall score is a weighted blend of four pillars, each on a 0 to 100 scale:
+ skillDurability × 0.25
+ roleTrajectory × 0.20
+ adjacencyBreadth × 0.20
Task Exposure is inverted because it measures how much of your work AI already touches: high exposure lowers readiness, so the engine counts what remains. The result is clamped to 0 to 100.
PILLAR 1 · 35%
Task Exposure
You pick the O*NET tasks you actually do. Each task carries an exposure value from observed AI usage in the Anthropic Economic Index; tasks without observed usage get a category-based estimate, and every value is labeled with its source. The pillar is the importance-weighted average of your selected tasks, so exposure on your core daily work counts more than exposure on rare tasks.
On your result, tasks at ≥ 60read as “AI is here”, ≤ 40 as durable, and the band between as shifting.
PILLAR 2 · 25%
Skill Durability
The weighted average durability of your role’s top 10 skills by importance, scored against automation-resistance categories and WEF skill trajectories. Roles whose skills are inherited from a parent occupation get a confidence discount rather than fabricated precision: × 0.95 for high-confidence inheritance, × 0.85 for moderate.
PILLAR 3 · 20%
Role Trajectory
Starts at a base of 50, then adjusts with three components: BLS 10-year projected growth (growth × 2, capped at ±30), the WEF industry outlook (net change × 0.5, capped at ±15), and an employment-scale buffer (log₁₀(jobs) × 5, capped at +30) because large occupations have hiring inertia. Missing data contributes zero and is flagged, never invented.
PILLAR 4 · 20%
Adjacency Breadth
Counts realistic escape routes: adjacent roles with skill overlap of at least 0.65, target durability of at least 60, and a retraining path under 36 months. Faster transitions score more: 10 points per role reachable under 12 months, 5 points for 12 to 24 months, 2 points for 24 to 36 months, capped at 100.
DATA
Sources, versioned.
O*NET 30.3
Task and skill structure for every occupation: what each role actually does, with importance weights.
Anthropic Economic Index (March 2026 release)
Observed AI usage per task, measured from real conversations. This is what "exposure" means here: observed usage, not speculation.
BLS Occupational Employment Statistics May 2025
Employment counts and 10-year projected growth per occupation.
WEF Future of Jobs Report 2025
Industry-level net role change outlook to 2030.
Every computed score stores the exact data versions it was computed against, alongside methodology version v1.0.0.
WHERE AI IS, AND ISN'T
LLMs never change scores.
Language models do three jobs around the score, never inside it: they match your job title to the right O*NET occupation, they sanity-check adjacent-role suggestions (filtering out transitions that ignore licensing or specialty boundaries), and they write the plain-language explanation of your result. If every one of those calls failed, your number would be identical.
HONESTY NOTES
Known limits.
Percentile comparisons describe people who took this assessment, not the workforce. Skill durability can under-differentiate rule-based entry-level roles. Exposure measures where AI is used, which is not the same as where jobs are lost; that is why the score blends exposure with durability, trajectory, and options rather than treating exposure as a verdict.
Questions or challenges are welcome: hello@jobroute.ai.